Article summary — original
Jane Street Announces Protocol Emulator ASIC Competition with Fabrication Funding
Jane Street has announced an open-source hardware competition challenging participants to design a general-purpose protocol emulator ASIC. Selected winning designs will be funded for fabrication via a Tiny Tapeout shuttle on IHP's 130nm process.
Participants must design an open-source protocol emulator capable of implementing protocols like UART, SPI, and I2C via reprogrammable firmware rather than fixed logic.
Targeting IHP's 130nm CMOS5L process via Tiny Tapeout, designs are capped at a 6x4 tile area (approx. 0.7 mm²), with a potential increase to 8x4.
Winners will receive fabricated chips mounted on development boards targeting the March 2027 CMOS5L shuttle.
Submissions are due by January 18, 2027, and team participation is strongly recommended.
Design Challenge and Capabilities
The competition requires designing a tiny CPU with an instruction set focused on reading/writing pins and precise cycle counting to emulate hardware protocols in software. Rather than incorporating fixed logic blocks for standard interfaces, the ASIC must remain reprogrammable after fabrication to accommodate new protocols under timing and I/O limits. Base targets are UART, SPI, and I2C, with stretch goals including low-speed USB and 10Mbit Ethernet, as well as interfaces such as JTAG, SWD, PS/2, and CAN bus.
Process Parameters and Resource Budget
The competition targets IHP's 130nm CMOS5L process using Tiny Tapeout tooling. The current design footprint is restricted to a 6x4 tile layout (24 tiles total, roughly 0.7 mm²), offering an estimated budget of about 1,000 logic cells per tile, though Jane Street is exploring scaling to 8x4 tiles. Participants are advised to leverage SRAM for instruction memory to conserve space, run synthesis and place-and-route early, and manage routing and clock-tree constraints.
Rules, Timeline, and Rewards
Submissions must be open source and submitted by January 18, 2027. Jane Street welcomes novel design and verification approaches, including formal methods, constrained random testing, and AI-assisted verification. Winning designs will be funded for fabrication on a Tiny Tapeout shuttle targeted for March 2027, with winners receiving mounted chips on development boards for silicon testing.
Saved original article
Last month, we asked you to
reverse engineer a
chip
from nothing but its
layout and teased a bigger challenge. Results and our favorite writeups are coming soon.
In the meantime, here’s our next challenge!
This time, you’re designing the chip, and we’ll pay to fabricate our favorite designs!
We’re particularly interested in projects with unique functionality, as well as those
that demonstrate novel approaches to design and verification methodologies! Winners will
receive a fabricated copy of their chip, mounted on a dev boards, so they can test their
design in real silicon.
The challenge
Design an
open-source, general-purpose protocol emulator ASIC
.
Hardware protocols like UART, SPI, and I2C are simple enough that people routinely
“bit-bang” them: toggle pins from software with careful timing instead of using a
dedicated peripheral. A protocol emulator is a small chip built to do exactly that: a tiny
CPU with an instruction set designed for reading pins, writing pins, counting cycles, and
hitting timing precisely enough that you can implement a real protocol in firmware rather
than in fixed logic. Something like that is a useful tool for hardware debugging and
reverse engineering, which is a good part of what we do.
The hard part is flexibility. The goal isn’t to put a UART block, an SPI block, and an
I2C block on one die and call it done. Your chip should be reprogrammable enough to support
new protocols
after
fabrication, within its timing and I/O constraints. For inspiration,
look at the PIO state machines on the RP2040 or the PRU cores on TI’s Sitara parts, and
consider what you’d do differently.
Start with UART, SPI, and I2C.
Stretch goals include low-speed USB and 10Mbit Ethernet.
Other interesting protocols to consider: JTAG, SWD, PS/2, CAN bus
If you have access to an FPGA, consider using it to test your RTL before the ASIC flow.
Show us anything else your architecture makes possible that we haven’t thought of.
At Jane Street, we use
Hardcaml
to generate the RTL for our FPGA
and ASIC designs. We are excited to see the languages and verification techniques you use,
including formal methods, random constrained tests, AI-assisted verification, and more.
As AI-assisted chip design becomes more common, we believe verification will be an
extremely important aspect of the ASIC design flow going forwards.
The rules
Process:
We’re targeting IHP’s 130nm CMOS5L process through our friends at
Tiny Tapeout
. Start with the
CMOS5L Verilog
template
, which
takes you from RTL to GDS. Set the tile size in
info.yaml
to 6x4.
Area:
The current maximum area is 6x4 tiles per design.
We are working on the
possibility of scaling up to 8x4 tiles (~30% more area). We’ll update this page, as well
as emailing everyone who has
signed
up
if the larger tile size becomes available.
Open source:
Your submission should be open source so others can use and build on it.
Unlike the reverse-engineering puzzle, there’s no need to keep your work hidden until
the deadline, so feel free to build in public!
Teams:
This is a much bigger project than the puzzle, so we strongly recommend
working in teams.
Deadline:
Submit your design by
January 18th, 2027
.
Prize:
We’ll pay to tape out the most novel designs on a Tiny Tapeout shuttle.
We’re targeting the
March 2027 CMOS5L shuttle
, subject to the foundry schedule.
Winners will receive chips and dev boards back after fabrication, so you can test your
design in silicon.
How much fits?
An 6x4 allocation is 24 tiles. At approximately 200um × 150um per tile, that’s about 0.7
mm² of nominal tile area. As a rough estimate, budget for about 1K logic cells per tile.
You may need to get creative to fit the functionality you want.
For instruction memory, SRAM can be more area-efficient than flip-flops. Tiny Tapeout has
examples of SRAM
running on this process node you can reference.
Run synthesis early, check the mapped cell area, and leave room for clock-tree buffers
and routing. Then run the full place-and-route flow and check timing. A design that looks
small enough after synthesis can still be difficult to route or too slow at your chosen
clock frequency.
Getting started
If you’ve never taped out a chip before, the
Tiny Tapeout documentation
walks through the process end
to end, and the tools are all free and open source. Start by getting a UART transmitter
out of a pin. Then make it programmable.
Sign-up
If you’re interested, please fill out our
sign-up
form
.
We’ll send updates about the tapeout template, deadlines, as well as providing the final
submission link. Note that filling out the form is not a commitment to participating, it’s
just to receive updates!
We’ll add a final submission form to this page closer to the deadline!
If you have questions along the way, reach out to
asic-competition@janestreet.com
.
Hardware at Jane Street
The hardware team at Jane Street designs FPGAs and ASICs that run some of the fastest
trading systems in the world. If designing a chip for fun sounds like your kind of thing,
take a look at our hardware
internships
and
full-time roles
. You can
also explore
Hardcaml
, our open-source OCaml hardware design
libraries, or
stay in touch
.